个人资料：

姓名：李蒙阳

职称：讲师、硕士生导师

通讯地址：天津师范大学电子与通信工程学院明理楼C区424

电子信箱：limengyang@tjnu.edu.cn

主要经历：

(1) 2025.02-至今 天津师范大学电子与通信工程学院，讲师、硕士生导师

(2) 2021.09-2024.11天津大学，深度学习与数据挖掘，博士

研究方向：

数据优化、偏好优化、具身智能

研究生招生：

目前计划招收2名硕士研究生，研究方向围绕数据优化、大语言模型、具身智能展开。欢迎对上述研究领域感兴趣、踏实认真、具备一定数学基础和编程能力（Python优先）的同学报考。有意者可通过上述电子信箱联系咨询，邮件请注明“硕士报考+姓名+本科院校/专业”。

论文发表：

[1]Mengyang Li,Haozhan Geng, Zhong Zhang, Shuang Liu. Dual-Difficulty Curriculum Learning for Direct Preference Optimization,KDD,2026,CCF A.

[2]Mengyang Li,Shuang Liu,Zhong Zhang.Layer-wise Gradient Disentanglement: Decoupling Semantics and Preferences in Direct Preference Optimization,ICML,2026,CCF A.

[3]Mengyang Li,Zhong Zhang, Pinlong Zhao. Task-Aware Preference Calibration for Direct Preference Optimization,ICML,2026,CCF A.

[4]Mengyang Li,Pinlong Zhao.DABO: Difficulty-aware Bayesian Optimization with Diffusion-learned Priors,CVPR Highlight,2026,CCF A.

[5]Mengyang Li,Jingwen Wang,Pinlong Zhao.What Do LLMs Learn First? Asymmetric Learning Dynamics of Input Complexity and Output Ambiguity in Preference Alignment,ACL,2026,CCF A.

[6]Mengyang Li,Pinlong Zhao,Zhong Zhang.Aligner, Diagnose Thyself: A Meta-learning Paradigm for Fusing Intrinsic Feedback in Preference Alignment,ICLR,2026,CCF A.

[7]Mengyang Li,Pinlong Zhao.Difficulty-Aware Learning Curve Extrapolation,AAAI,2026,CCF A.

[8]Mengyang Li,Xiaoling Zhou, Ou Wu.Delving into the Training Dynamics for Image Classification,IEEE TIP，2025，CCF A，中科院一区TOP，IF=13.7.

[9]Mengyang Li,Weiyao Zhu.Ou Wu,Toward Learnable and Interpretable Data Shapley Valuation for Deep Learning,KBS，2025，中科院一区TOP，IF=7.2.

[10]Mengyang Li, Fengguang Su, Ou Wu, Ji Zhang.Class-level Logit Perturbation,IEEE TNNLS，2024，CCFB，中科院一区TOP，IF=8.9.

[11]Mengyang Li, Fengguang Su, Ou Wu. Logit Perturbation,AAAI，2022，CCF A.

[12]Mengyang Li,Xudong Zhou,Pinlong Zhao.Learning Temporally-Aware Sample Weights for Preference Optimization,ACL Findings,2026.

[13]Mengyang Li, Ou Wu. ECOC-IL: Robust and Efficient Label LDP for Imbalanced Learning,CVPR Findings, 2026.

[14]Pinlong_Zhao,Huijun Tang,Pengfei Jiao,Mengyang Li(通讯).What Tokens Truly Matter? The Logit Conflation Problem in LLM Sampling,ACL Findings,2026.

[15]Ou Wu,Mengyang Li.Revisiting the Effective Number Theory for Imbalanced Learning, IEEE TKDE，2024，CCF A，IF=8.9.

[16]Pinlong Zhao,Mengyang Li, Pengfei Jiao, Huijun Tang, Ou Wu, Unequal Vulnerability: The Differential Impact of Label Flipping Attacks Across Classes,WWW,2026,CCF A.

[17]Xiaolin Zhou, Ou Wu,Mengyang Li.Investigating the Sample Weighting Mechanism Using an InterpretableWeighting Framework, IEEE TKDE, 2023, CCF A，IF=8.9.

[18]Ou Wu, Tao Yang,Mengyang Li, Ming Li.Two-Level LSTM for Sentiment Analysis With Lexicon Embedding and Polar Flipping, IEEE Transactions on Cybernetics，2022，中科院一区TOP，IF=11.8.

[19]Yu Zhu; Yingchun Ye,Mengyang Li, Ji Zhang, Ou Wu.Investigating annotation noise for named entityrecognition. 2023, IF=6.0.

专利发表：

[1]一种票据类图像版面解析方法及装置，发明授权。

[2]一种面向内容安全的困难样本挖掘方法及装置，发明授权。

[3]一种可解释性的深度学习样本赋权方法，发明公开。

[4]一种面向财务报表图像的科目信息提取方法，发明公开。

[5]一种面向财务报表图像的结构化信息提取方法，发明公开。